Alexey Pospelov is the Machine Learning team leader at the Search Quality department of Yandex LLC in Moscow, Russia. Yandex is the most popular search engine in Russia handling over 60% of the country's search traffic. Alexey joined Yandex in 2012 as a search quality developer. Previously, he worked from 2005 to 2009 as a software engineer and later as a research scientist in the domains of the VLSI logic validation and computer microarchitecture at Intel Corp. He made his PhD (Russian equivalent) in Algebraic Complexity Theory at Moscow State University in 2008. From 2009 to 2012 he was a postdoc at the Computational Complexity group at Saarland University.

AG 1, AG 2, AG 3, AG 4, AG 5, RG1, SWS, MMCI   Info

MPI Audience

English

60 Minutes

Saarbrücken

The problem of ranking the documents according to their relevance to a given query is one of the principal issues of industrial Information Retrieval. Most learning-to-rank methods are supervised and use human editor judgements for learning. In this talk I will give a general overview of the way Yandex ranks the web using heuristic and machine-learned algorithms. I'll stop on the Matrixnet machine learning technology, which is an open algorithm with a proprietary implementation sitting at the heart of the Yandex Search. I'll focus on the differences of Matrixnet from the traditional gradient boosted trees, its uses within Yandex and its competitive advantages against other industrial solutions.

Klaus Berberich

--email hidden

passcode not visible

logged in users only